Hi all. Interesting article and the appeal of a "set it and forget it" mechanical system is always tempting for traders, but of course, one has to be very skeptical when establishing a system based on one year's worth of data. For example, how does 135 minutes as an opening range perform over 10 or 20 years of testing? Why is taking a position one tick above or below such a range statistically significant? Does the market 'care' that prices went up or down a tick at this interval? If 10 years of data shows this to be significant then there's something to consider. But I've read of other statistical analyses, notably by Mark Fisher in the Logical Trader whereby a significant percentage of daily ranges can be predicted based on the first 15 minutes of trading as opposed to the 135 minute parameter set forth in this article. I bring this up not as a criticism but to point out the potential landmines one might encounter. In any case, I did perform a backtest of the year 2018 using the article's parameters, EXCEPT not trading on Thursday or avoiding days where the previous day's range was 1.6x greater than it's previous 5 days; I didn't think these 2 parameters would be ultimate deal breakers.
In any case, the 2018 results were interesting. 216 trades were taken, 121 win, 92 loss, total gross = 220,400 minus 17,210 commissions for a net profit of $203,190. The fixed ratio/delta was used for leveraging contracts.
On its face, this is an obviously great result. But then I tested year 2015 and looked at several months of tests from randomly chosen years between 2008 and 2012, and the results were horrible, with steady losses, and large daily draw downs. Thus, the 2018 results really tell us nothing other than, IF a trader had chosen THAT particular year to utilize the system, he/she would have made tremendous profit. But if the system were used in, say 2015 or in other select years, losses would have been significant. It is doubtful a trader would have continued for very long
I tried adding some filters including VIX as a counter-trend signal, $TRIN (for example not long taking trades if $TRIN was => 1.1), changed time frames, used 5 minute charts to look for breakouts then pullbacks, pre-assigned stop losses, etc., and nothing elicited any edge. So on this account, I fully agree with the article author that indicators, or any 'tweaking' of the system, would prove futile.
If anyone has any additional ideas/input on opening range breakout models as per this article, I welcome them. The basic jist of trading an ORB system remains fascinating if for no other reason than the mechanical nature of such systems often appears - at least in theory - to potentially supersede losses, by removing emotions from the trade.